Graph coloring algorithms for multi-core and massively multithreaded architectures

We explore the interplay between architectures and algorithm design in the context of shared-memory platforms and a specific graph problem of central importance in scientific and high-performance computing, distance-1 graph coloring. We introduce two different kinds of multithreaded heuristic algorithms for the stated, NP-hard, problem. The first algorithm relies on speculation and iteration, and is suitable for any shared-memory system. The second algorithm uses dataflow principles, and is targeted at the non-conventional, massively multithreaded Cray XMT system. We study the performance of the algorithms on the Cray XMT and two multi-core systems, Sun Niagara 2 and Intel Nehalem. Together, the three systems represent a spectrum of multithreading capabilities and memory structure. As testbed, we use synthetically generated large-scale graphs carefully chosen to cover a wide range of input types. The results show that the algorithms have scalable runtime performance and use nearly the same number of colors as the underlying serial algorithm, which in turn is effective in practice. The study provides insight into the design of high performance algorithms for irregular problems on many-core architectures.

► We explore the interplay between architectures and algorithm design for graph problems.
► We target multi-core platforms and a massively multithreaded system.
► We present two different kinds of multithreaded algorithms for graph coloring.
► The implementations are evaluated on three platforms (Nehalem, Niagara 2, Cray XMT).
► Scalable performance on each platform is demonstrated.